Closed Derek-Jones closed 3 years ago
Thanks for the suggestion, despite this very slow reply. We will leave the constant at the current level of precision for now. The method of computing CMP probabilities - which is also used to truncate series in the normalizing constant - should be overhauled in the not-too-distant future.
The C++ function cmp_allprobs needsh more digits of accuracy in its Euler–Mascheroni constant: 0.57721566490153286060651209008240243104215933593992